Guard: A Guided AI System for Intrusion Detection And Automated Response In Critical Infrastructure Environments

An anomaly-based IDS using Large Language Models was developed by a team of four within a three-week time frame. The initiative commenced on July 25th, with the initial week dedicated to evaluating the research papers, sources, and existing code examples. The task was to implement the idea in a way that would encompass the supply of a fully operational IDS. Within the next three weeks, we developed Shell scripts in python to effectively capture and preprocess captured network packet data. This preprocessed data would be fed into an IDS to identify potentially suspicious activity. Empirical data indicated that the system had the capability to detect anomalies in the network traffic. Thereby, proving its value for enhancing the security controls through an IDS based on a language Model. The present study presents the potential for the augmentation of LLM-based solutions within the domain of intrusion detection.